Quantitative proteomics reveal temporal proteomic changes in ...

Article pubs.acs.org/jpr

Quantitative Proteomics Reveals Temporal Proteomic Changes in Signaling Pathways during BV2 Mouse Microglial Cell Activation Jongmin Woo,†,∥ Dohyun Han,‡,§,∥ Joseph Injae Wang,‡ Joonho Park,‡ Hyunsoo Kim,‡ and Youngsoo Kim*,†,‡ †

Department of Biomedical Sciences and ‡Department of Biomedical Engineering, Seoul National University College of Medicine, 103 Daehangro, Seoul 110-799, Korea § Proteomics Core Facility, Biomedical Research Institute, Seoul National University Hospital, 101 Daehangro, Seoul 110-799, Korea S Supporting Information *

ABSTRACT: The development of systematic proteomic quantification techniques in systems biology research has enabled one to perform an in-depth analysis of cellular systems. We have developed a systematic proteomic approach that encompasses the spectrum from global to targeted analysis on a single platform. We have applied this technique to an activated microglia cell system to examine changes in the intracellular and extracellular proteomes. Microglia become activated when their homeostatic microenvironment is disrupted. There are varying degrees of microglial activation, and we chose to focus on the proinflammatory reactive state that is induced by exposure to such stimuli as lipopolysaccharide (LPS) and interferon-gamma (IFN-γ). Using an improved shotgun proteomics approach, we identified 5497 proteins in the whole-cell proteome and 4938 proteins in the secretome that were associated with the activation of BV2 mouse microglia by LPS or IFN-γ. Of the differentially expressed proteins in stimulated microglia, we classified pathways that were related to immune-inflammatory responses and metabolism. Our label-free parallel reaction monitoring (PRM) approach made it possible to comprehensively measure the hyper-multiplex quantitative value of each protein by highresolution mass spectrometry. Over 450 peptides that corresponded to pathway proteins and direct or indirect interactors via the STRING database were quantified by label-free PRM in a single run. Moreover, we performed a longitudinal quantification of secreted proteins during microglial activation, in which neurotoxic molecules that mediate neuronal cell loss in the brain are released. These data suggest that latent pathways that are associated with neurodegenerative diseases can be discovered by constructing and analyzing a pathway network model of proteins. Furthermore, this systematic quantification platform has tremendous potential for applications in large-scale targeted analyses. The proteomics data for discovery and label-free PRM analysis have been deposited to the ProteomeXchange Consortium with identifiers and , respectively. KEYWORDS: BV2 microglial activation, dimethyl labeling quantification, quantitative proteomics, multiple reaction monitoring−mass spectrometry (MRM-MS), parallel reaction monitoring (PRM)

■

INTRODUCTION

Moreover, microglial activation-mediated inflammatory responses normally cause neuronal damage and remove damaged cells through phagocytosis.5,6 Activated microglia express various cytokines and growth factors in response to nerve injuries under pathological conditions.7 To observe the active states of intracellular and intercellular microglia, two major inflammationassociated molecules, lipopolysaccharide (LPS) and interferongamma (IFN-γ), have been used to stimulate BV2 mouse microglial cells. LPS induces dopaminergic neurodegeneration through NF-kB and Toll-like receptor signaling.8,9 IFN-γ activates microbicidal effector functions through JAK-STAT signaling and the enhancement of innate immune responses.10

Microglia are cellular sentinels of innate immunity that are critical in the development of neurodegenerative diseases, the prevalence of which has risen worldwide as the average life expectancy has increased steadily. According to a recent study, microglia help maintain immune surveillance by responding quickly to pathogens that enter the parenchyma when the central nervous system (CNS) is compromised by foreign invaders.1 Microglia have many functions, including antigen presentation and processing,2 activation of T cells and B cells,3 and initiation of nonspecific immune responses following stimulation. Microglia undergo a series of morphological changes in their active state. As part of this process, various surface receptors are upregulated to create a variety of biochemical repertoires that help in different subtypes of the immune reaction.4 © XXXX American Chemical Society

Received: June 23, 2017 Published: August 4, 2017 A

DOI: 10.1021/acs.jproteome.7b00445 J. Proteome Res. XXXX, XXX, XXX−XXX

Article

Journal of Proteome Research LPS and IFN-γ synergize to amplify the expression of nitric oxide synthase (iNOS), which increases nitric oxide (NO) production.11,12 The microglial active state has been examined merely with regard to a few factors to several hundred associated molecules13−16 and has not been analyzed on a proteomic scale that takes into account the related pathway networks. Systems biology is the study of the organization of dynamic and complex biological components that are difficult to predict from individual characteristics. To study these components, systematic measurement techniques, such as genomics, bioinformatics, and proteomics, have been used to quantify groups of interactors. In particular, quantitative proteomics has evolved to incorporate the in-depth analysis of a cellular system.17,18 Many studies have attempted to identify specific proteins using a quantitative proteomics approach.19−21 However, conventional targeted methods, such selected reaction monitoring (SRM)based targeted proteomics, are labor-intensive and inconvenient because separate platforms are used for the discovery and targeted analysis.22 Moreover, one is restricted to analyzing 100−200 target peptides, rendering this approach unsuitable for larger-scale applications.23 To overcome these limitations, we designed a systematic proteomics approach that can track changes in more proteins on a single platform by mass spectrometry compared with current methods. Several quantitative mass-spectrometry-based technologies have recently been developed, of which isotopic labeling after sampling provides a variety of options for quantitative analysis because it is applicable to multiple samples.24,25 In particular, dimethyl-labeling-based quantification at the MS1 level allows the ion intensities of peptide pairs or triplets to be compared simultaneously.26 Stable-isotope dimethyl labeling has several advantages compared with other multiplex labeling approaches such as stable isotope labeling with amino acids in cell culture (SILAC): It can tag nearly any biological sample without any limitations to the choice of protease, uses inexpensive reagents, and labels reagents quickly and keeps them relatively stable. However, its drawbacks are that it can introduce variations into the early steps during sample preparation when performed on digested peptides of those samples and that it broadly profiles a proteome by data-dependent acquisition (DDA), which has limited quantitative application due to missing values in replicated.26,27 Thus DDA-based analysis is unsuitable based on its inability to detect low-abundance proteins. To mitigate the disadvantages of DDA-based MS acquisitions, a targeted proteomics analysis is required to accurately quantify proteins that are identified by global proteome profiling. On a high-resolution quadrupole orbitrap, parallel reaction monitoring (PRM) can acquire full MS/MS spectra with high specificity and separate coisolated interference ions from target peptides with a tolerance of approximately 10−20 ppm.28 Technically, PRM can be performed with or without labels. A label-based method can be used to determine the absolute and relative levels of proteins in samples by spiking them with heavy isotopelabeled synthetic peptides, analogous to targeted endogenous peptides. In contrast, the label-free method is straightforward and suitable for hypermultiplexed targeted proteomics for semiquantitative measurements.29 For label-free PRM approaches, single or multiple heavy isotope-labeled reference peptides (LRPs) can help normalize the peak areas of all endogenous target peptides, assess the performance of the system, and correct for variations in retention time shifts between LC runs.30,31

In this study, dimethyl-labeled proteomics and label-free PRM were used to develop a systematic proteomics approach that allows simultaneous and targeted quantification on a highresolution Q-Exactive orbitrap MS platform. Using this method, we examined the dynamic changes in proteins during microglial activation, induced by a single effector or a combination of factors as model systems. Over 450 peptides that were derived from the target proteins were quantified simultaneously in a single run by label-free targeted analysis. Recent technological and methodological advances have made the acquisition of >500 peptides in a single LC−MS run possible on quadrupole Orbitrap28 and QqTOF instruments.32 Our approach differs in that it enables one to quantify over 450 peptides in a single run by label-free targeted analysis. Thus systematic proteomics identifies intracellular and intercellular proteins, the expression of which is regulated significantly during activation. On the basis of these data, the construction and analysis of a pathway network model of proteins can reveal latent pathways that are associated with neurodegenerative diseases.

■

EXPERIMENTAL SECTION

Cell Line Culture and Treatments

BV2 mouse microglial cells were cultured in DMEM, containing 10% FBS and 1% penicillin and streptomycin. A 100 mm dish was seeded with 1 × 106 cells. After a 24 h incubation at 37 °C with 5% CO2, serum deprivation was performed with serum-free media, and the cells were allowed to adjust for at least 4 h. In the control sample, pellets were scraped and washed three times with 1× DPBS, and the conditioned media were collected and concentrated on 3 kDa conical filters for 24 h (Amicon Ultra, Merck Millipore, Darmstadt, Germany) to a volume of ∼200 μL.33 BV2 cells were plated in a 100 mm dish and exposed to one of two stimuli for 24 h: 1 μg/mL Escherichia coli LPS or 10 ng/mL IFN-γ that was dissolved in serum-free DMEM. In the case of the targeted analysis stage, BV2 cells were activated with 1 μg/mL LPS, 10 ng/mL IFN-γ, or 1 μg/mL LPS plus 10 ng/mL IFN-γ for 6, 12, 24, and 48 h. At each time point, the conditioned media and pellet were harvested as in the control sample. All treatments were performed in triplicate. Protein Isolation and Digestion

Cell pellets were lysed at room temperature in lysis buffer (4% SDS, 1 mM DTT, 0.1 M Tris-Cl, pH7.4) with 1 min of sonication. Conditioned media was added to the lysis buffer (4% SDS, 2 mM DTT, 0.1 M Tris-Cl), and the lysates were boiled and centrifuged at 15 000 rpm for 10 min. The concentration of the proteins in the lysates was measured using a BCA assay kit (Thermo Fisher Scientific, Rockford, IL). Then, 200 μg of the protein sample was digested by filter-aided sample preparation (FASP) with slight modifications.34 The proteins were loaded onto a 30 K spin filter (Millipore, Billerica, MA), and the buffer was exchanged with UA solution (8 M urea in 0.1 M TrisCl, pH 8.5) by centrifugation. After triple UA exchange, the reduced cysteines were alkylated with 0.05 M iodoacetamide (IAA) in UA solution for 30 min at room temperature in a dark room. We exchanged the UA buffer to 40 mM ammonium bicarbonate (ABC) and digested the samples with trypsin (enzyme-to-substrate ratio of 1:100) at 37 °C for 18 h. For of the construction library and the targeted analysis, the digested peptides were acidified with trifluoroacetic acid and cleaned using homemade C18 Stage-Tip columns. B

DOI: 10.1021/acs.jproteome.7b00445 J. Proteome Res. XXXX, XXX, XXX−XXX

Article

Journal of Proteome Research Stable-Isotope Dimethyl Labeling and Cleaning of Peptides

PRM mode with an isolation width of 2 m/z, a target AGC value of 1.0 × 106, and a maximum injection time of 100 ms. To optimize the PRM analyses, the PRM event used a resolution of 35 000 or 70 000 on the orbitrap. Ion activation/dissociation was performed with a higher-energy c-trap dissociation of 25, 27, or 30. For regular PRM analyses, the PRM event used a resolving power of 35 000 with a normalized collision energy of 27 on the orbitrap. PRM scans were triggered by scheduled targeting precursor ions that were selected for endogenous peptides in ±5 and ±1.5 min elution windows for the evaluation of detectability and the targeting method, respectively

Dimethyl triplex labeling was performed according to standard protocols with several modifications.26 Digested peptides were loaded onto a 200 μL of homemade C18 Stage-Tip column that was packed with POROS 20 R2 resin. The peptides were then tagged with stable-isotope dimethyl labels comprising three mixtures: regular formaldehyde and cyanoborohydride (28 Da shift, designated “light label”), deuterated formaldehyde and regular cyanoborohydride (32 Da shift, designated “intermediate label”), and deuterated and 13C-labeled formaldehyde with cyanoborodeuteride (36 Da shift, designated “heavy label”). We flowed the dimethyl labels intermittently through the column five times for 10 min. The labeled peptides were then washed with loading buffer that contained 0.1% TFA. Cleaned peptides were eluted and dried on a vacuum concentrator. Three biological replicates were established for each treatment group: 2 for the “forward” experiment and 1 for the “reverse” experiment. For the forward experiment, untreated control peptides were labeled with the light label, whereas peptides from cells that were stimulated with LPS or LPS+IFN-γ were labeled with the intermediate or heavy label, respectively. For the reverse experiment, LPS-treated samples were tagged with heavy dimethyl labels, and the control and LPS+IFN-γ-treated samples were tagged with light and intermediate dimethyl labels, respectively.

Raw Data Search

Raw MS files for the dimethyl-labeled quantification were processed using the Andromeda search engine, the built-in peptide identification algorithm in Maxquant (ver. 1.3.0.5), to match the spectra against the UniProtKB FASTA database (74 540 entries, version from June 2014). MS/MS searches were performed with the following parameters: peptide length of at least six amino acids, fixed carbamidomethylation modification, variable methionine oxidation, and variable N-terminal acetylation. The tolerance was set to 6 and 20 ppm for main search and first search, respectively. A false discovery rate (FDR) of 1% was applied to all protein and peptide searches. The retention times of all analyzed samples were linearized with the “Match between runs” feature of Maxquant, which allows identified peptides to be transferred in the absence of sequencing with a retention window of 2 min. Only for the dimethyl-labeled quantification were three multiplicities selected with stable-isotope dimethyl labeling (control: +28 Da, intermediate: +32 Da, heavy: +36 Da) at the lysine and N-terminus. All proteins were filtered for common contaminants, such as keratins. All proteomic data sets for dimethyl-labeled quantification were submitted to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange. org/cgi/GetDataset) through the PRIDE partner repository with identifier .36

Peptide Fractionation with High-pH Reverse-Phase Fractionation

Dimethyl-labeled peptides were fractionated per an established StageTip-based high-pH reverse-phase protocol with some modifications.35 Dimethyl-labeled peptides and desalted peptides were resolved in 200 μL of loading buffer (10 mM ammonium formate, pH 10 and 2% acetonitrile) and loaded onto 200 μL yellow tips that were packed with C18 Empore disk membranes (3 M, Bracknell, U.K.) and POROS 20 R2 resin (Invitrogen, Carlsbad, CA). Loaded peptides were eluted with ACN gradient buffer solution at pH 10 in 13 fractions. The fractions were dried on a speed vacuum concentrator and stored at −80 °C pending the LC−MS analysis.

Prediction Tools for Putative Secreted Proteins

Proteins that were detected in conditioned media were filtered to predict secreted proteins using SignalP 4.1 (http://www.cbs.dtu. dk/services/SignalP/), SecretomeP 2.0 (http://www.cbs.dtu.dk/ services/SecretomeP/), and the TMHMM server 2.0 (http:// www.cbs.dtu.dk/services/TMHMM/) in comparison with the Exocarta database (http://www.exocarta.org). The FASTA forms of the identified proteins were exported and used for this analysis. The resulting protein lists were submitted to the SignalP 4.1 server. “Eukaryotic organism group” and a cutoff value of 0.45 were used as prediction parameters. The SecretomeP 2.0 server was used to determine whether the proteins were nonclassically secreted, with the organism “mammal” and a threshold score of 0.5 as parameters. The TMHMM server 2.0 was used with the FASTA forms to predict transmembrane helices in the proteins. The resulting data set was compared with exosomal proteins from Exocarta.

Mass Spectrometric Analysis of Dimethyl Labeling

Peptides were separated on a Nanoflow Easy-nLC 1000 (Proxeon Biosystems, Odense, Denmark) that was equipped with a trap column (100 Å, 3 μm particle, 75 μm × 2 cm) and an analytical column (100 Å, 1.8 μm particle, 50 mm × 15 cm). A gradient that ranged from 5 to 30% acetonitrile was run at a fixed flow rate of 300 nL/min for 180 and 60 min for the relative quantitation and targeted analysis, respectively. Eluted peptides were ionized at the tip of the column at a spray voltage of 2.00 kV. Ionized peptides were analyzed on a quadrupole-orbitrap mass spectrometer (Q-Exactive, Thermo Fisher Scientific, San Jose, CA). MS1 spectra were measured in DDA mode at a resolution of 70 000 with an m/z range of 300 to 1800 and a target automatic gain control (AGC) of 3.0 × 106 with 100 ms maximum fill times. The 20 most abundant ions were selected with an isolation window of 2 m/z, fragmented by higher-energy collisional dissociation (HCD) with a normalized collision energy of 30 and a resolution of 17 500 at 100 m/z. The dynamic exclusion of sequenced peptides was fixed to 30 s to restrict repeated sequencing. Each sample was analyzed in triplicate for technical replicates.

Target Protein Selection and in Silico Fragmentation

Target selections in the whole-cell proteome (WCP) and secretome (SEC) groups were processed as shown in Supplementary Figure S3A. Normalized ratios from Maxquant were filtered statistically in two steps: (1) Two paired groups were compared by their significance B values, which were calculated for the protein subsets that were obtained, based on intensity, with a Benjamini−Hochberg FDR < 5%. (2) ANOVA with p-value